Speaker identification in smart environments with multilayer perceptron
Jasmina Novakovic · 2011
This paper presents reliability of MLP in speaker identification using characteristics extracted from their voices. Classification accuracy depends on speaking condition and varies up to 23% depending on the selected speaking condition. Results of simulation experiment show that MLP is effective in speaker identification, especially in the case of retelling and synchronous speech where we achieved a recognition rate of max 80.25%. In smart environment, identification by speech can be sufficiently accurate, expecially when it is combine with others characterics of a person such as face, behavior, actions, etc.